Enhancing Email Security: A Hybrid Machine Learning Approach for Spam and Malware Detection

Abstract

Recent research indicates a notable surge in SMS spam, posing as entities aiming to deceive individuals into divulging private account or identity details, commonly termed “phishing” or "email spam". Conventional spam filters struggle to adequately identify these malicious emails, leading to challenges for both consumers and businesses engaged in online transactions. Addressing this issue presents a significant learning challenge. While initially appearing as a straightforward text classification problem, the classification process is complicated by the striking similarity between spam and legitimate emails. In this study, we introduce a novel method named "filter" designed specifically for detecting deceptive SMS spam. By incorporating features tailored to expose the deceptive techniques employed to dupe users, we achieved an accurate classification rate of over 99.01% for SMS spam emails, while maintaining a low false positive rate. These results were attained using a dataset comprising 746 instances of spam and 4822 instances of legitimate emails. The filter's accuracy, evaluated on a dataset with two attributes and 5568 instances, notably surpasses existing methodologies. Our proposed model, a Hybrid NB-ANN model, achieves the highest accuracy at 99.01%, outperforming both Naïve Bayes (98.57%) and Artificial Neural Network (98.12%). This highlights the efficacy of the hybrid approach in enhancing accuracy for email spam detection and malware filtering, ensuring comprehensive coverage across training and test datasets for improved feedback loops.